#5535: Make clients use "a" lines in network status documents
-------------------------+--------------------------------------------------
Reporter: ln5 | Owner: ln5
Type: enhancement | Status: accepted
Priority: normal | Milestone: Tor: 0.2.4.x-final
Component: Tor Client | Version:
Keywords: ipv6 | Parent: #4564
Points: | Actualpoints:
-------------------------+--------------------------------------------------
Changes (by ln5):
* cc: arma (added)
Comment:
> > When should a client prefer the IPv6 or port of a (non bridge) entry
> node?
>
> My first thought would be either "never" or "only when told to do so" at
first. My rationale is that lots of clients will have IPv6 addresses that
leak their MAC addresses for a while, and we don't (now) have code to
detect that yet.
>
> >Let's try this: Prefer IPv6 if there is an IPv6 OR port but make sure
we fall back to its IPv4 port if we fail connecting.
>
> I'm not saying that's wrong, but I'd like to hear your rationale.
I'm afraid my rationale is based more on functionality and
connectivity than MAC protection. It boils down to two points:
- Using IPv6 would (supposedly) be good for connectivity when running
on networks without IPv6 blocking.
- I have become less worried about MAC address leaking lately, mostly
because my (highly nonscientific) list of systems with privacy
extensions for SLAAC (RFC 4941) enabled by default now includes most
of what we care about (Ubuntu, Debian, Fedora, OS X and Windows).
I have to admit that the value of the first point and the risk
mitigation given by the last point are both hard to quantify.
New proposed solution is to revisit #4455 and expand the discussion
beyond bridges, taking MAC privacy into account.
--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/5535#comment:9>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online